Admission Control in Time-Slotted Multihop Mobile Networks

Transcription

1 dmission ontrol in Time-Slotted Multihop Mobile Networks Shagun Dusad and nshul Khandelwal Information Networks Laboratory Department of Electrical Engineering Indian Institute of Technology - ombay Mumbai India bstract The emergence of nomadic applications have recently generated a lot of interest in next-generation wireless network infrastructures which provide differentiated service classes. So it is important to study how the quality of service (QoS), such as packet loss and bandwidth, should be guaranteed. To accomplish this, we study an admission control scheme which can guarantee bandwidth for real-time applications in multihop mobile networks. We examine via simulation the system performance in various QoS traffic flows and mobility environments. On-demand of the protocol feature enhances the performance in the mobile environment because the source can keep more connectivity with enough bandwidth to a receiver in the path-finding duration. Simulation experiments show this improvement. Introduction Mobile d-hoc Networks (MNETs) are future wireless networks consisting entirely of mobile nodes that communicate on-the-move without base stations. Nodes in these networks will both generate user and application traffic and carry out network control and routing protocols. Rapidly changing connectivity, network partitions, higher error rates, collision interference, and bandwidth and power constraints together pose new problems in network control, particularly in the design of higher level protocols such as routing and in implementing applications with Quality of Service requirements. Without an inherent infrastructure, the mobiles handle the necessary control and networking tasks by themselves, generally through the use of distributed control algorithms. Multihop connections, whereby intermediate nodes send the packets toward their final destination, are supported to allow for efficient wireless communication between parties that are relatively far apart. d hoc wireless networks are highly appealing for many reasons. They can be rapidly deployed and reconfigured. They can be tailored to specific applications. They are also highly robust due to their distributed nature, node redundancy, and the lack of single points of failure. Personal communications and the mobile computing require a wireless network infrastructure that is fast deployable, possibly multihop, and capable of multimedia service support. The wireless network is often connected to a wired network (e.g., TM or Internet) so that the TM or Internet multimedia connection can be extended to the mobile users. There are several contributions that have already appeared in the wireless extensions of the wired TM networks. Most of them focus on the cellular architecture for wireless personal communication networks (PNs) supported by TM backbone infrastructures. In this architecture, all mobile hosts in a communication cell can reach a base station in one hop. The problem of interconnecting the multihop wireless network to the wired backbone requires a quality-of-service (QoS) guarantee not only over a single hop, but also over an entire wireless multihop path. The key to the support of QoS reporting is QoS routing, which provides path QoS (bandwidth) information at each source. Prior research efforts in multihop mobile networks have not fully addressed this problem.. Problem Formulation We address the problem of supporting realtime communications in a multihop adhoc network using QoS routing, and we study a protocol for QoS routing. We consider different QoS traffic flows in the network to evaluate the performance of our protocol. Multimedia applications such as digital audio and video have much more stringent QoS requirements than traditional datagram applications. For a network to deliver QoS guarantees, it must reserve and control resources. major challenge in multihop, multimedia networks is the ability to account for resources so that bandwidth reservations (in a deterministic or statistical sense) can be placed on them. We note that in cellular (single hop) networks, such accountability is made easily by the fact that all stations learn of each others requirements, either directly or through a con-

2 trol station (e.g., the base station in cellular systems). However, this solution cannot be extended to the multihop environment. To support QoS for real-time applications, we need to know not only the minimal delay path to the destination, but also the available bandwidth on it. V (Virtual onnection) should be accepted only if there is enough available bandwidth. Otherwise, it would disrupt the existing V.s. We only consider bandwidth as the QoS (thus omitting signal-to-interference ratio (SIR), packet loss rate, etc.). This is because bandwidth guarantee is one of the most critical requirements for real-time applications. andwidth in timeslotted network systems is measured in terms of the number of free slots. The goal of the QoS routing algorithm is to find a shortest path such that the available bandwidth on the path is above the minimal requirement. To compute the bandwidth-constrained shortest path, we not only have to know the available bandwidth on each link along the path, but we also have to determine the scheduling of free slots. There are two types of schemes complimented with DM used in mobile adhoc network M protocol : TDM based [] Here we use time slots for data transfer and control information transfer. lso different nodes can use different codes so even adjacent nodes can transfer data in the same slots. The main constraint here is the synchronization of the time slots which still needs overheads (in the form of preamble part of packets). pplying highly synchronized solutions in an ad hoc network becomes expensive and synchronization can fail when the nodes are mobile. The slot allocation algorithm in TDM schemes is also vulnerable to mobility in the network since slot allocations must be reconfigured whenever there are changes in available bandwidth or changes to routes in the network. RTS/TS based SM/ [] Here we use RTS/ TS signals to reserve the bandwidth. The bandwidth depends not only on the node but also the nodes inside the carrier sensing region of the node in view. This is because if one node is transmitting then no node in the sensing region can transmit.. Organization of the Report Section. discusses the bandwidth calculation algorithm giving details of the proposed system model. Section. describes the On-demand nature of the protocol. Section describes the simulation results. Finally, we conclude in Section. System Model The system consits of arbitrary distributed nodes following addmission control based on [].. andwidth alculation Lin and Liu [] proposed a new bandwidth routing scheme which contains bandwidth calculation and reservation for mobile ad hoc networks. In this protocol, the bandwidth information is embedded in the routing table. y exchanging the routing table, the end-to-end bandwidth of the shortest hop-distance path for a given source-destination pair can be calculated. If there is not enough bandwidth over the shortest path, the call request will be blocked. However, not enough bandwidth over the shortest path does not mean that there does not exist any bandwidth route in the network. That is, there may be a route which meets the bandwidth requirement but is not the shortest in hop distance. Therefore, this protocol may miss some feasible bandwidth routes and the blocking probability is high. In this protocol, we would like to route and reserve resources for a connection in a way that: ) minimizes the blocking probability by attempting several routes in parallel; ) always considers the feasible route with the minimal cost (the shortest route); ) selects a route with a cost that is close to the min-cost feasible route; and ) limits the flow of information and the use of computing resources in that process. ontrol Phase Data Phase Slot Figure : TDM time frame structure. ontributions of this work Our main contribution is to simulate the given protocol []. This protocol is based on TDM/DM. We simulate a network where the nodes are following this protocol and measure the effective throughput, average number of calls, average number of incomplete calls against the mobility of nodes and the average arrival rate for various QoS requirements. The admission control protocol [] used here is Ondemand type. Figure : No collision at within DM system s was the network environment discussed in [], the transmission time scale is organized in frames, each containing a fixed number of time slots. The entire network is D

3 synchronized on a frame and slot basis. Namely, time is divided into slots, which are grouped into frames. Propagation delays will cause imprecision in slot synchronization. However, slot guard times (fractions of microsecond) will amply absorb propagation delay effects (in the order of microseconds). Each time a frame is divided into two phases: a control phase and data phase. The size of each slot in the control phase is much smaller than the one in the data phase. The TDM time frame structure is shown in Fig. The control phase is used to perform all the control functions, such as slot and frame synchronization, power measurement, code assignment, V setup, slots request, etc. The amount of data slots per frame assigned to a V is determined according to the bandwidth requirement. We assume a TDM within our network; a code division multiple access (DM) is overlaid on top of the TDM infrastructure. Multiple sessions can share a common TDM slot via DM. In this case, the near-far problem and related power control algorithm become critical to the efficiency of the channel access. n ideal code assignment scheme [] is assumed running in the lower layer of our system, and all spreading codes are completely orthogonal to each other. Thus the hidden terminal problem can be avoided. onsider the example illustrated in Fig.. can use the same slots as to send packets to D encoded by a different code without any collision at. It is notable that this case is assumed only one session through,, and D. If and (different sessions) intend to send packets to in the same slot, then only one packet can be received and another will be lost depending on which code locks on. one hop distance link bandwidth from to One or more hop distance Path bandwidth from to alculation of path bandwidth from to Figure : adwidth calculation overview ecause only adjacent nodes can hear the reservation information, and the network is a multihop, the free slots recorded at every node may be different. We define the set of the common free slots between two adjacent nodes to be the link bandwidth. onsider the example shown in Fig. in which intends to compute the bandwidth to. We assume the next hop is. y using our end-to-end bandwidth calculation scheme, if can compute the available bandwidth to, then can use this information and the link bandwidth to to compute the bandwidth to. We define the path bandwidth (also called end-to-end bandwidth) between two nodes, which are not necessarily adjacent, to be the set of available slots between them. If two nodes are adjacent, the path bandwidth is the link bandwidth. We can observe that link W (P, Q) = freeslot(p ) freeslot(q). freeslot(x)is defined to be the slots, which are not used by any adjacent host of X to receive or to send packets, from the point of view at node X. Next, we can further employ link bandwidth to compute the end-to-end bandwidth. This information can provide us an indication of whether there is enough bandwidth on a given route between a source-destination pair. We consider four different cases which may arise while calculating the end-to-end bandwidth. Figure : The Equal case Figure : The containing case case : In this case, (refer Fig. ) just the floor of the size of anyone one of the bandwidth divided by gives the path bandwidth size of to. ase : In this case, (refer Fig. ) the minimum of the two sets, one obtained by subtracting the subset from superset and the subset itself gives the path bandwidth size from to. ase : In this case, (refer Fig. ) both the bandwidth are mutually exclusive so just the minimum of the two gives the path bandwidth size from to. ase : This is a general case as shown in Fig.. We can successively calculate the effective bandwidth by applying all the three cases listed above as this is the combination of them. The steps for reducing this to the above cases are shown in Fig. and Fig. 9 Figure : The exclusive case

4 9 Figure : The general case 9 Figure : Step 9 Figure : Overview of the On-demand routing Figure 9: Step In general, to compute the available bandwidth for a path in a time-slotted network, one not only needs to know the available bandwidth on the links along the path, but also needs to determine the scheduling of the free slots. Resolving slot scheduling at the same time as available bandwidth is searched on the entire path is a NP-complete problem. We assume a simple heuristic based algorithm to achieve an suboptimal solution.. On Demand Routing In the case of multihop networking, most routing protocols for packet radio networks can be categorized as being before-demand or on-demand protocols. efore-demand protocols compute and maintain routes even if nodes are not actively transmitting packets. Generally, each host needs to maintain a distance vector-based routing table. In contrast, on-demand protocols compute routes only when necessary. n over view of the on-demand routing used here is illustrated in Fig.. In [], the shortest path is the only candidate when searching for a feasible bandwidth route. ecause of before-demand bandwidth calculation, a host can decide either to accept or to reject a new call immediately without any delay. Due to the rapidly changing availability of resources and the processing delay, it is difficult and impractical to use the before-demand routing approach to maintain the pool of candidates for each source-destination pair. On-demand routing can also save the control messages for maintaining inactive routes. ecause it is on-demand, there will be delay for the virtual circuit setup but we ignore this delay for the sake of simplicity in the simulation set up. Simulation Results We simulate the proposed admission control scheme considering the environment which consists of mobile hosts roaming uniformly in meter square area. Each node moves randomly at uniform speed. The transmission range is m. In our model, effect of error caused by the channel variation is ignored. we have paid more attention to the effect of mobility upon the system performance. data slots are assumed to be in data phase. Since the number of data slots is less than the number of nodes, nodes need to compete for these data slots. The source-destination pair of a call is randomly chosen and their distance must be greater than one. Once a call request is accepted on each of the nodes present in the route, only then data-slots are reserved on each of the nodes. The slots are released when either the session is finished or the connection is broken between any two immediate nodes in the route. We have simulated for average number of connections and average throughput against the mobility of users. We observe from Figure and that for both the cases as the arrival rate increases the throughput of the system also increase implying that it has not reached to its capacity. We observe that for the case with QoS requirement of four slots (refer Fig. ) the system reaches its capacity for higher values of arrival rate whereas it could not reach to its full capacity with QoS requirement of slots (Fig. ). lso, we observe that failure rate increases with increase mobility when the system performs near its capacity (refer Fig. ) onclusion In summary, we have simulated an admission control over an on-demand routing protocol which is suitable for use

5 Figure : verage number of incomplete calls versus call arrival rate for mobility m/s and QOS slots 9 9 Figure : verage throughput versus call arrival rate for mobility m/s and QOS slots Figure : verage number of incomplete calls versus call arrival rate for mobility m/s and QOS slots Figure : verage throughput versus call arrival rate for mobility m/s and QOS slots Figure : verage number of calls versus call arrival rate for mobility m/s and QOS slots with multihop mobile networks. Here the route chosen needs not to be the shortest one. ut it must satisfy the QoS Figure : verage number of calls versus call arrival rate for mobility m/s and QOS slots requirement. Therefore we need to find all the paths from source to destination. This admission control can be applied to two important scenarios: multimedia ad-hoc wire-

6 ..... Figure : verage number of incomplete calls versus mobility for mean interarrival time slots and QOS slots Figure : verage number of incomplete calls versus mobility for mean interarrival time slots and QOS slots Figure : verage throughput versus mobility for mean interarrival time slots and QOS slots Figure : verage throughput versus mobility for mean interarrival time slots and QOS slots Figure 9: verage number of calls versus mobility for mean interarrival time slots and QOS slots Figure : verage number of calls versus mobility for mean interarrival time slots and QOS slots less networks and multihop extension wireless TM networks. Specially, the bandwidth information can be used to assist in performing the handoff of a mobile host between two TM base stations. The on-demand routing helps in the

7 networks where the due to the rapidly changing availability of resources and the processing delay, it is difficult and impractical to use the before-demand routing approach to maintain the pool of candidates for each source-destination pair. Therefore, here we consider on-demand routing protocol which not only can search for several routes in parallel for a connection, but also can incorporate our bandwidth calculation scheme. References []. R. Lin, dmission control in timeslotted multihop mobile networks, IEEE Journal on Selected reas in ommunications vol. 9, no. : Oct, pp.9-9. [] Y. Yang and R. Kravets, ontention-aware admission control for adhoc networks, UIU Tech Report,. []. R. Lin and J. S. Liu, QoS routing in ad hoc wireless networks, IEEE J. Select. reas ommun., vol., pp.., ug. 999.